To address the fragmentation of AI agent integrations, OpenAI, Microsoft, Amazon, Vercel, and Anysphere have released "Agent Plugins," an open-standard specification version 1.0 designed to enable cross-platform interoperability for agent extensions. Initiated by Vercel, the standard aims to replace disparate folder layouts and setup requirements with a unified package format. Each plugin consists of a directory containing a root plugin.json file that bundles two core components: Model Context Protocol (MCP) servers, which facilitate connections to live data and tools, and "Agent Skills," which are modular, reusable instruction sets. This "build once, run anywhere" approach is supported at launch by ChatGPT, Codex, Cursor, GitHub Copilot, Kiro, and VS Code. The specification intentionally remains narrow, focusing strictly on packaging and discovery while delegating complex issues—such as sandboxing, permissions, installation, and security verification—to individual clients. While critics argue the standard is too thin to solve deep integration challenges, proponents suggest it will democratize the ecosystem by allowing small developers to reach multiple major agent platforms simultaneously. The steering committee, composed of the five primary backers, maintains that the project is openly licensed and independent of any single company's product roadmap.